15:58:35 <shelikhoo> #startmeeting tor anti-censorship meeting
15:58:35 <shelikhoo> here is our meeting pad: https://pad.riseup.net/p/tor-anti-censorship-keep
15:58:35 <shelikhoo> feel free to add what you've been working on and put items on the agenda
15:58:35 <MeetBot> Meeting started Thu Oct 13 15:58:35 2022 UTC.  The chair is shelikhoo. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:58:35 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:58:39 <shelikhoo> Hi!
15:58:51 <itchyonion> hello
15:59:04 <cohosh> hi
16:06:31 <shelikhoo> I didn't see any new topic...
16:06:36 <shelikhoo> other than one
16:06:50 <shelikhoo> about Release a new version of snowflake webext proxy
16:07:29 <shelikhoo> if you are adding things to the upper part, I might miss it
16:08:34 <shelikhoo> okay, let begin the first topic
16:08:34 <shelikhoo> Release a new version of snowflake webext proxy
16:08:55 <shelikhoo> we have merged a few merge request in snowflake webext
16:09:05 <shelikhoo> it might be a good time to release a new version
16:09:14 <shelikhoo> any blockers?
16:09:23 <cohosh> yes
16:09:40 <cohosh> there's still some work to be done on getting translations working smoothly again
16:10:14 <cohosh> right now we have to manually pick which translations to update
16:10:24 <cohosh> i'm hoping this will be resolved in a week or two
16:10:50 <shelikhoo> yes, we could wait for a few weeks...
16:11:18 <shelikhoo> let's discuss this again in next weeks meeting and see whether we could move forward
16:11:36 <shelikhoo> Okay next topic is
16:11:37 <shelikhoo> loss of bandwidth at snowflake-01 bridge
16:11:37 <shelikhoo> dynamics are the same as at the time of 2022-10-06 meeting
16:11:37 <shelikhoo> the cause of the loss of bridge bandwidth is still unknown https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowflake/-/issues/40207 https://gitlab.torproject.org/tpo/anti-censorship/team/-/issues/96#note_2840481
16:12:24 <dcf1> I don't have anything to add other than what's in the links
16:14:35 <shelikhoo> If this issue is caused by server
16:14:43 <shelikhoo> rather than the proxy part
16:14:55 <shelikhoo> we could fix this by distributed snowflake proxy
16:15:21 <shelikhoo> I will begin the process to roll out the distributed snowflake proxy support to Tor browser
16:15:36 <shelikhoo> with the 02.snowflake.torproject.net server
16:15:38 <dcf1> well, I think it must be more complicated than that
16:16:07 <dcf1> but a second bridge is something we need to move ahead with anyway
16:16:40 <shelikhoo> yes, so we can remove a set of hypnosis at no extra cost to us
16:16:48 <dcf1> one thing I'm concerned about is that the hardware for snowflake-02 is not quite as fast as the hardware for snowflake-01
16:17:14 <shelikhoo> This should be fine, it is not going to receive a lot of users in the beginning
16:17:15 <shelikhoo> and
16:17:23 <dcf1> I think it's a good idea to push out support first in the alpha release only, in order to gauge capacity.
16:17:48 <dcf1> snowflake-02 is able to handle 50% of the users we have today, but probably not 50% of the users we had 10 days ago.
16:18:07 <shelikhoo> we would be able setup even more bridges without additional cost
16:18:30 <dcf1> other than coordination, maintenance, communication costs
16:19:08 <dcf1> I just saw a link from cohosh in #tor-dev: https://gitlab.torproject.org/tpo/core/tor/-/issues/40636#note_2819763
16:19:30 <shelikhoo> yes, I think the first step is to make it possible to use the second bridge with a custom bridge line
16:19:44 <shelikhoo> we would be able setup even more bridges without additional coding(fixed)
16:19:46 <dcf1> Probably unrelated, but I was not previously aware that snowflake-01's high usership might be interpreted by other relays as a DoS signal
16:21:12 <cohosh> linked to me, i don't know how to tell if this is due to snowflake's load balancing or if it indicates another issue
16:22:53 <cohosh> they seemed surprised at the number of connections even given the 12 instances, but that machine is seeing a lot of traffic
16:24:04 <dcf1> Just by pigeonhole principle, the snowflake bridge is going to be making (much) more than one connection to some relays, since there are many times more snowflake users than there are relays
16:24:48 * cohosh nods
16:25:11 <dcf1> https://share.riseup.net/#JWQAl-2UMIyak9FeP6OJ8A even back at the level of 20k
16:26:00 <dcf1> this goes with something I posted in my notes
16:26:02 <dcf1> https://gitlab.torproject.org/tpo/network-health/metrics/website/-/issues/40009#note_2841814
16:26:18 <dcf1> https://gitlab.torproject.org/tpo/anti-censorship/team/-/wikis/Research-ideas#calibrate-bridge-users-estimation-with-on-bridge-socket-counts
16:27:13 <dcf1> I suspect that the Tor Metrics estimation algorithm is inflating user counts, not hugely, maybe by about a factor of 2 at most
16:27:41 <dcf1> at this moment the number of open sockets is
16:27:42 <dcf1> # ss -n state established 'dst 127.0.0.1' 'dport = 10000' | wc -l
16:27:42 <dcf1> 10656
16:28:22 <dcf1> but still, it's a lot of connections funneled through one IP address
16:29:30 <cohosh> yeah, wow
16:30:28 <shelikhoo> I think snowflake bridge should be given an exception when it comes to DoS protection, since it wouldn't work if that restriction is applied to it strictly
16:30:32 <shelikhoo> in its current form
16:31:06 <shelikhoo> but the good thing is that that restriction is not currently respected...
16:31:08 <dcf1> yeah I was wondering if maybe something changed with respect to DoS parameters in the consensus, or something like that, at 2022-10-04 17:15
16:31:22 <dcf1> but it still wouldn't explain the contemporaneous drop in client polling
16:31:52 <dcf1> which is why I rather suspect something to do with blocking of the broker than with something on the bridge or something to do with the Tor network
16:32:05 <dcf1> (though that still doesn't explain why not only Iran had a drop at that time)
16:34:42 <shelikhoo> There was a broker deployment that rejected old proxies... but there was a difference in time between two events
16:35:06 <shelikhoo> however, they are close enough I am unsure how timezone come into play
16:35:22 <dcf1> I thought about that possibility, but the time difference was more than 24 hours
16:36:00 <dcf1> I don't know, is it worth re-deploying the broker not to reject old proxies, just as a test, for 48 hours or so?
16:36:55 <cohosh> that sounds like a reasonable idea to me
16:37:12 <shelikhoo> okay, I will do this next Monday
16:37:39 <dcf1> still I'm baffled why it should have affected client polling rate
16:37:41 <dcf1> https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowflake/-/issues/40207#note_2840780
16:37:58 <dcf1> I guess I shouldn't let one piece of contradictory evidence override other pieces of contradictory evidence :)
16:38:41 <dcf1> It's possible, like, 2 different things happened near the same time and the effects can't be explained by a single cause
16:39:26 <dcf1> the only other avenue of investigation I had in mind is to run a proxy and try to see if the DTLS connections are getting interfered with
16:40:05 <cohosh> i agree this is really weird
16:40:24 <cohosh> a very sudden drop that doesn't align with a deployment *exactly* that affected everywhere at once
16:40:47 <shelikhoo> yes... anyway i will try to revert the change on broker next monday]
16:40:54 <cohosh> thanks shelikhoo
16:41:06 <shelikhoo> because there is only one way to find out....
16:41:45 <shelikhoo> anything more on this topic?
16:41:52 <shelikhoo> now is the announcement time
16:41:53 <shelikhoo> snowflake-01 transferred 1.17 PB of Tor user data in the six months between 2022-04-08 and 2022-10-08
16:41:53 <shelikhoo> users with annotated events https://share.riseup.net/#JWQAl-2UMIyak9FeP6OJ8A
16:41:53 <shelikhoo> bandwidth https://share.riseup.net/#qpzsTgHxfG0Ar-JrDRfQEg
16:42:07 <shelikhoo> ===================
16:42:08 <shelikhoo> New release v1.1.3 of uTLS
16:42:08 <shelikhoo> https://github.com/refraction-networking/utls/releases/tag/v1.1.3
16:42:08 <shelikhoo> supports more and more recent TLS parrots
16:42:08 <shelikhoo> from https://github.com/net4people/bbs/issues/129#issuecomment-1276774330
16:42:13 <shelikhoo> ===============
16:42:34 <shelikhoo> do we wants to discuss about any of the announcements?
16:45:00 <shelikhoo> okay, anything else we wants to discuss in this meeting?
16:45:36 <cohosh> (awesome about the announcments)
16:45:40 <cohosh> but no more to add from me
16:47:21 <shelikhoo> okay, let's call it a meeting! Thanks everyone~~~
16:47:22 <shelikhoo> #endmeeting